#emc-devel | Logs for 2008-12-06

[02:01:11] <SWPadnos> it would be great for emc to have its own mesasge log, even if DEBUG isn't set all that high
[02:01:35] <SWPadnos> so you could go back and get messages in situations like Stuart's
[03:06:17] <SWPadnos> oh hey - I found a .tar.gz of the ez-trol code
[03:06:42] <cradek> I thought I sent that to you a while back
[03:07:52] <cradek> sadly, stuart's operator reset emc and the bad behavior went away
[03:08:00] <SWPadnos> mmm
[03:08:02] <cradek> prior to that, it was repeatably doing it.
[03:08:08] <SWPadnos> bummer
[03:08:19] <cradek> yeah. I was ready to go down there to see it.
[03:08:38] <SWPadnos> I thought there had been an ez-trol email, but I may have been looking for it in the wrong folders
[03:09:11] <cradek> I could fix it if I could see it. I've been after it for two years now I bet.
[03:09:12] <SWPadnos> I don't suppose you've tried to reproduce it in a mangled pseudo-dalih sim config?
[03:09:16] <SWPadnos> yeah
[03:09:20] <cradek> yes I definitely tried
[03:10:14] <cradek> that's why I asked for the config
[03:10:50] <cradek> seb_kuzminsky: does coverity say anything about emccanon.cc?
[03:11:05] <SWPadnos> the screenshot was very interesting. it didn't just ignore the arcs and make them straight lines. it combined them with the prior straight lines and made a single motion to the arc endpoint
[03:11:19] <cradek> nope you're looking at it backwards
[03:11:25] <SWPadnos> oh :)
[03:11:34] <cradek> it skips the arc, then the next line goes FROM the wrong point to the programmed endpoint
[03:11:39] <SWPadnos> ah, ok
[03:11:43] <SWPadnos> damn! :)
[03:12:49] <cradek> dangit dangit dangit I wish he had not reset it
[03:13:19] <cradek> it was compiled there so I could easily have attached the debugger and seen wtf is wrong
[03:13:25] <SWPadnos> I guess someone could have suggested turning up the debug level and trying to run again
[03:13:35] <SWPadnos> run the part that is
[03:13:45] <SWPadnos> I didn't think of it when emailing
[03:13:58] <cradek> I bet the arc 'task issue' calls would be missing, but I agree it would be great to verify that.
[03:17:04] <seb_kuzminsky> cradek: http://highlab.com/~seb/cov/query-run-table.html
[03:17:20] <seb_kuzminsky> that's the list of coverity issues from TRUNK as of last night
[03:17:35] <seb_kuzminsky> nothing in canon
[03:17:40] <cradek> oh it's shorter than I figured
[03:17:43] <cradek> ok, thanks
[03:17:46] <cradek> it was a long shot.
[03:18:06] <seb_kuzminsky> there's several real bugs there, possibly related, i dont know
[03:46:05] <SWPadnos> hmmm. bummer. the lathe at HGR is on sale today for $1249 on eBay
[03:46:33] <SWPadnos> someone might buy it before jmkasunich gets a chance to look at it and see if it's worth the $800 they want for it in the store :)
[03:47:18] <cradek> SWPadnos: can't you be there yourself in about 8 hours?
[03:47:31] <SWPadnos> sure, but they're closed on Saturdays
[03:47:52] <SWPadnos> or is it that there's a special sale one saturday but they're still open the other times
[03:49:18] <cradek> I think they're closed 3/4 saturdays
[03:49:39] <SWPadnos> that's what I thought too, and I think the special saturday this month is the 20th
[03:49:46] <jmkasunich> they are open one saturday a month
[03:49:55] <jmkasunich> 3rd one I think
[03:50:17] <jmkasunich> I will be on vacation Monday the 15th, FWIW
[03:50:28] <SWPadnos> cool. will you be at HGR? :)
[03:50:34] <jmkasunich> good chance
[03:50:41] <jmkasunich> I missed last month
[03:51:46] <SWPadnos> It's probably worth $800, it doesn't look like any parts blew up or anything
[03:52:03] <SWPadnos> no evidence of major chemical fires on the outside
[03:52:10] <jmkasunich> worth 800 to someone anyway
[03:52:17] <jmkasunich> the question is, worth 800 to you?
[03:52:23] <SWPadnos> heh
[03:52:37] <SWPadnos> if the motors and mechanicals are in good shape, then probably
[03:52:47] <cradek> that's the $800 question
[03:52:52] <SWPadnos> drives can be found, controls are easy
[03:53:04] <SWPadnos> shipping is a PITA though :)
[03:53:24] <SWPadnos> someone has a trailer on eBay that has a swing-out crane on it
[03:53:43] <SWPadnos> but it can't handle more than 1000-2000 pounds (the crane anyway)
[03:54:02] <jmkasunich> if you have a trailer, loading is easy (they'll do it)
[03:54:05] <cradek> that lathe isn't much over 2000 lb I bet
[03:54:12] <jmkasunich> unloading is harder - rent a forklift, etc
[03:54:25] <jmkasunich> you have more time to unload, unless you need to return the trailier right away
[03:54:33] <SWPadnos> yep - it's the last 3 feet (vertical) that make it hard
[03:54:38] <cradek> tell me about it
[03:54:47] <cradek> 1 foot
[03:54:49] <SWPadnos> I can borrow a trailer from a friends dad actually, so that's not a problem
[03:54:51] <SWPadnos> heh
[03:54:58] <SWPadnos> 3 inches canbe hard with a 3000 pound machine :)
[03:55:30] <cradek> that's a stack-up-some-wood height
[03:55:35] <jmkasunich> 3" isn't hard with patience and a prybar
[03:55:40] <cradek> right
[03:55:40] <jmkasunich> yeah, what cradek said
[03:55:44] <SWPadnos> heh
[03:55:58] <SWPadnos> it's definitely doable
[03:56:17] <SWPadnos> we had fun getting my friends lathe into his basement workshop (walk-out, fortunately)
[03:56:40] <SWPadnos> a 3000-pound Takisawa, suspended on a tow truck boom (that had non-functional hydraulics)
[03:56:44] <SWPadnos> (at midnight)
[03:57:13] <SWPadnos> through a doorway, then a small entryway, then another doorway - both "standard" 30" or so width
[05:01:55] <SWPadnos> hmmm.
[05:02:20] <SWPadnos> so I added a G540 presets button to stepconf, and it sets some invert pins for step and estop I/Os
[05:02:44] <SWPadnos> should I make the Xylotex and Sherline preset buttons clear those?
[05:07:34] <SWPadnos> oh hmmm. stepconf doesn't notice if you select the same function for multiple inputs
[05:11:42] <CIA-42> EMC: 03cmorley 07v2_2_branch * 10emc2/src/hal/classicladder/ (classicladder.c classicladder_gtk.c classicladder_gtk.h): Changes to have the section window display the ladder program name-backport from CL 8.2
[05:14:11] <seb_kuzminsky> autoconf is automatically confusing me
[05:14:19] <CIA-42> EMC: 03cmorley 07TRUNK * 10emc2/src/hal/classicladder/ (classicladder.c classicladder_gtk.c classicladder_gtk.h): Changes to have the section window display the ladder program name-backport from CL 8.2
[05:15:01] <cradek> seb_kuzminsky: uh-oh, what ar eyou trying to do?
[05:15:26] <SWPadnos> GNU autopandemonium
[05:16:02] <seb_kuzminsky> i'm trying to add check to our stuff <http://check.sf.net>
[05:17:08] <seb_kuzminsky> the non-realtime stuff can use it
[05:17:19] <seb_kuzminsky> ah there it goes
[05:18:06] <seb_kuzminsky> usually when projects use autoconf, they also use automake too, right? but we dont
[05:18:28] <cradek> I bet often they do - not sure about usually
[05:59:35] <CIA-42> EMC: 03cmorley 07TRUNK * 10emc2/src/hal/classicladder/drawing.c: change background colour of section display...If everyone hates it I will change back...
[06:05:20] <CIA-42> EMC: 03cmorley 07TRUNK * 10emc2/src/hal/classicladder/zEMC_README.txt: update
[07:01:53] <seb_kuzminsky> no, they're arguments? really?
[07:13:19] <seb_kuzminsky> what does it mean for a private member to be static?
[07:13:54] <seb_kuzminsky> as in: emc/rs274ngc/rs274ngc.hh:490 static setup _setup;
[07:16:16] <SWPadnos> http://www.icce.rug.nl/documents/cplusplus/cplusplus10.html
[07:16:46] <SWPadnos> it gets initialized only once for the class, not for each instance of the class (I guess)
[07:16:56] <seb_kuzminsky> ah, it's a "class variable"
[07:16:58] <SWPadnos> but it's private data that only the class can access
[07:17:00] <SWPadnos> yes
[07:17:05] <SWPadnos> that's how I read that document
[07:17:08] <seb_kuzminsky> thx
[07:17:14] <SWPadnos> sure
[07:17:37] <SWPadnos> amazing actually
[07:17:55] <SWPadnos> I searched gogle for "static private member" and got no porn hits
[07:19:16] <SWPadnos> night
[07:27:42] <seb_kuzminsky> goodnight SWPadnos
[11:45:28] <micges> how can I pass arguments to rt hal module (kinematics) ?
[12:24:30] <alex_joni> you make parameters
[12:24:37] <alex_joni> and use setp to change those parameters
[13:43:53] <micges> I mean when loadrt
[13:44:01] <micges> like count=1
[14:09:20] <micges> to choose which axis is rotational and which is flat
[14:16:21] <jepler> micges: read the manpage rtapi_module_param
[14:16:25] <jepler> then grep for some examples
[14:18:35] <micges> jepler: thanks
[14:27:26] <jepler> * jepler kicks CIA-42
[14:27:32] <CIA-42> ow
[16:17:26] <Guest945> Guest945 is now known as skunkworks
[16:26:37] <micges> what is reason that screw compensation is build in to emc ?
[16:27:03] <micges> not in kinematics
[16:40:30] <jepler> backlash compensation existed before hal
[16:41:38] <micges> oh ok
[16:41:42] <jepler> that's my guess anyway
[16:41:51] <jepler> "why" questions are hard to answer
[16:47:54] <alex_joni> even screw comp was available before hal
[16:59:14] <micges> while I'm learning what kinematics can do I'll try to move code for those corrections and see what will be happening
[17:00:05] <micges> next whole week I'll have 1.50x2.50m table mill for testing and developing new ideas
[17:02:18] <micges> in a 3 month we will add automatic tool changer to it
[17:02:55] <micges> will se how 2.2 will be able to control that
[17:05:36] <micges> bbl
[18:01:41] <SWPadnos> kinematics and actuator compensation are logically separate, though related
[18:02:01] <SWPadnos> kinematics is the conversion between differnt mathematical models of the world
[18:02:06] <jmkasunich> screw comp is a single-axis thing, skew comp involves multiple axes
[18:02:22] <jmkasunich> so screw comp can be handled at a lower level
[18:02:25] <SWPadnos> screw comp (and backlash) are intended to compensate for errors in the mechanical system
[18:02:48] <SWPadnos> right - good point, comp operates on errors in a single joint
[18:37:43] <micges> so now mechanical system errors are compensated in emc, and kinematics are pure mathematic conversion
[18:44:54] <alex_joni> you can compensate for sytem errors in kinematics too
[18:45:09] <alex_joni> especially if the system errors involve two or more joints
[18:46:42] <fenn> might work better for screw comp too, since you can use i.e. sine function to correct for periodic error
[18:51:35] <micges> are there any other mechanical system error that can be fixed with math ? (screw, skew, backlash)
[18:52:22] <fenn> it depends on the particular machine
[18:52:58] <fenn> nobody's done it yet with emc, but you should be able to fix a non-straight axis
[18:53:12] <micges> mill and lasers
[18:54:34] <micges> fenn: you mean skew ?
[18:55:50] <fenn> no, i mean if your rail isnt straight
[18:56:05] <fenn> it might be curved
[18:56:08] <fenn> or twisted
[18:57:21] <micges> I see, I have one machine with that
[18:58:24] <micges> what is general idea with curved rails ?
[18:58:31] <micges> control points ?
[18:59:37] <fenn> i have no idea
[19:00:12] <fenn> a spline might work
[19:16:09] <alex_joni> depends how it's curved
[19:16:16] <alex_joni> might be a simple arc
[19:16:26] <alex_joni> or something higher degrees
[19:19:54] <micges> I have rail folded in one place
[19:20:13] <micges> simplest example I think
[19:43:17] <alex_joni> folded?
[19:44:45] <micges> straight to 1m, a little curve and stright to end
[19:47:42] <skunkworks> as I think SWPadnos pointed out - unless you have a 5 axis machine - kins cannot correct the mill/beam parallel/perpendicularness to the axis.
[19:51:07] <micges> ok
[19:51:10] <micges> good night